Online Convex Optimization 2: Regularization
The goal of an OCO algorithm is to minimize the regret unlike traditional convex optimization problems.
This motivates a family of algorithm called Regularized Follow the Leader. This is based on a simple idea
of taking the optimal decision on the hindsight. Formally,
xt+1=x∈Kargminτ=1∑tfτ(x)
This is called Follow the Leader strategy in Machine Learning. This strategy fails miserably in worst case scenario.
Consider the formulation K=[−1,1],Let,f1(x)=21x,f2(x)=−21x and
Let fτ for τ=2,⋯,T alternate between −x and x. Thus,
τ=1∑tfτ(x)={21x, t is odd−21x, otherwise
The FTL strategy keeps shifting between xt=1 and xt=−1, always making the wrong choice. The algorithm can be modified to make
it more stable. Such means of stablization is called regularization.
1. Regularization Functions
Def 1: A function f is called α- strongly convex if
f(y)≥f(x)+∇f(x)T(y−x)+2α∣∣y−x∣∣2
Def 2: A function f is called β- smooth if
f(y)≤f(x)+∇f(x)T(y−x)+2β∣∣y−x∣∣2
Regularization functions R:K→R, are strongly convex and smooth.